A Genome-Wide Association Scan for Episodic Memory Decline in Aging

Gene discovery in Alzheimer's disease (AD) has largely relied on the case/control study design; however, this approach is limited by heterogeneity of dementia in cases and subclinical disease in controls. We have implemented a complementary strategy using episodic memory decline as a continuous outcome for genome-wide association (GWA) analysis. GWA analysis was performed using 2.5 million imputed single nucleotide polymorphism (SNP) genotypes and up to 15 years of prospective clinical data from the Religious Orders Study and Rush Memory and Aging Project (n = 1,593). The outcome phenotype was a composite quantitative measure of episodic memory decline slope, based on longitudinal assessment of 7 cognitive tests, and adjusted for age, gender, and education. For replication, we evaluated the top results using a similar memory decline trait in the Chicago Health and Aging Project (CHAP) (n = 588), and additionally tested for direct associations with AD susceptibility in the Cohorts for Heart and Aging Research Epidemiology (CHARGE [1,367 cases / 11,931 controls]). Among our results, rs17668591, within the alpha-dystrobrevin gene (DTNA), was associated with episodic memory decline in the discovery (p = 3x10−5) and CHAP replication cohorts (p = 0.01), and showed improved significance in a joint analysis (p = 2x10−6). We also discover rs11833579 (p = 9x10−5) within ninjurin2 (NINJ2), a locus that was previously associated with ischemic stroke and has been implicated in neuronal injury responses. In addition to memory decline, NINJ2 is associated with AD susceptibility (p = 0.001), and we find evidence for an enhanced, interaction effect in APOE epsilon-4 allele carriers (OR = 2.8, p = 1x10−4, pinteraction = 0.002). The association between NINJ2 and AD is further replicated in CHARGE (p = 0.02). Finally, our study reveals strong association signals with episodic memory decline at each of the validated AD susceptibility loci: APOE (p = 5x10−15), CR1 (p = 0.003), CLU (p = 0.002), PICALM (p = 2x10−4), and BIN1 (p = 3x10−4). Our strategy identifies DTNA, NINJ2, and several other genes as novel candidate risk factors for AD, and along with the known loci, reveals a broader impact on susceptibility for memory decline within the aging population.